algorithm - A step-by-step set of instructions or procedures for performing a task or solving a problem.  Examples include recipes, mathematical proofs, and computer programs.

artificial general intelligence (AGI) - A type of artificial intelligence capable of learning from and applying knowledge flexibly to any task that humans can perform. This type of AI is currently theoretical. Also known as STRONG AI.

artificial intelligence (AI) - The theory behind, study of, or development of hardware devices or computer programs that can perform tasks that typically require a human brain.

artificial superintelligence - A type of artificial intelligence that surpasses the human brain’s capabilities in most tasks; e.g. Data from Star Trek. This type of AI is currently theoretical.

deep learning - A type of machine learning that uses neural networks with at least three layers. These networks are very good at discovering and classifying unlabeled data.

diffusion model - A type of generative AI primarily used in video and image generation, as well as computer vision tasks. It operates through a process called iterative denoising, in which an image is gradually refined from random noise by repeatedly removing noise and learning patterns in the data. This technique enables the generation of highly detailed and realistic images from seemingly chaotic starting points.

generative adversial network (GAN) - An early and influential generative AI model that employs two neural networks, the generator and the discriminator, in a competitive framework. The generator aims to produce realistic content, while the discriminator evaluates the content to distinguish between real and generated data. Through this adversarial process, the generator continually improves its ability to create convincing outputs, and the discriminator becomes better at identifying authenticity. GANs are widely used in applications such as natural language processing, computer vision, and text and image generation, and remain a foundational approach in generative AI.

generative artificial intelligence (GAI) - A type of artificial intelligence that learns from giant diverse datasets with minimal human assistance to generate new content based on probability.

large language model (LLM) - A type of neural network designed to understand and replicate text based on what the computer program has learned about types of human writing from training on very large datasets.

machine learning - A subset of AI that uses algorithms and statistical models to enable computer systems to iteratively learn how to perform specific tasks typically completed by humans without being directly programmed to do so.

narrow AI  - A type of artificial intelligence designed to perform a specific task or narrow range of tasks.  Currently, all forms of available artificial intelligence are narrow AI. Aka WEAK AI.

neural networks - A subset of artificial intelligence that relies on large numbers of processors, known as nodes, connected in a way that imitates the human brain. Typically, neural networks learn by recognizing patterns in the training data provided to them rather than being explicitly programmed to process particular data in a particular way.

neural radiance fields (NeRFs) - A type of generative AI, pioneered by NVIDIA, that leverages deep learning to generate highly detailed 3D representations of objects or scenes from 2D images. NeRFs work by modeling the color and light density at any point in a 3D space, enabling the reconstruction of complex shapes and textures with photorealistic accuracy. This technology has applications in areas such as computer graphics, virtual reality, and digital content creation.

retrieval augmented generation (RAG) - A method to improve the accuracy and quality of LLM output through retrieving reliable curated information from an external database that the LLM can then check its output against before providing the output to the user.

supervised machine learning - A type of machine learning where the computer model is provided with labeled data (i.e., data that is annotated with what it is and/or the correct answer to a question) that it can use to find patterns and structures to use to perform its assigned task.

training data - Extremely large datasets (in the trillions of documents) often developed by scraping the Internet that are used to train artificial intelligences.

unsupervised machine learning - A type of machine learning where the computer model is provided with raw, unlabeled data that it can use to find patterns and structures to use to perform its assigned tasks.